Change my mind: pseudogroups are the "wrong way" to formalize differential geometry.
What's wrong with formalizing charts and atlases, then a manifold is a set equipped with a maximal atlas?
We could formalize, e.g., complex manifolds using G-structures.
What is wrong with this approach?